Thoughts on charity
I wrote this in February 2012. Some of the details have changed since then, but overall it remains a good summary of my thinking on these issues. I mention GiveDirectly in the text below without passing much judgement on them; while most of my donations still go to the Against Malaria Foundation, I now rate GiveDirectly, and cash transfers more generally, very highly (see my guest post at the GiveWell blog on cash transfers and deworming).
I am not an expert on charities. But as an interested layman, I have spent hundreds of hours thinking about the topic over the last two and a half years. In the following essay, I try to systematically describe my thoughts and feelings about giving to charity. Despite its length (5700 words), it is not comprehensive. In some areas I could go into more detail, and at least one topic has been omitted because it is of low importance and outside the main flow of reasoning.
I don't really have a target audience in mind, except that I wrote this in part for my future self – I would like a record of my thoughts and feelings on charity (I've already forgotten some of the development of my thinking), and in 2015 I'll be able to look back at this essay and remember how I felt in 2012. I am occasionally asked for my opinions on where or how to donate; it is probably useless directing curious people to an essay this long, but perhaps some excerpts will be useful for me or others to copy-paste in discussions on the topic. I also know that I occasionally do a blog search for GiveWell, and temporarily lurk in forum threads and blogs of like-minded people; maybe there are other people with this strange habit, who might enjoy the following tour of my mind. Lastly, I fondly hope that the essay will be of general interest.
TL;DR: Donate to the Against Malaria Foundation because GiveWell recommends them, and give 10% of your salary because Giving What We Can says so.
'It's the force that drove Theresa, and that school that Oprah built; I'll give you fifty bucks to take away my guilt.' – Tim Minchin (NSFW lyrics).
I don't agree with many of the details in Tim Minchin's highly amusing Guilt Song, but I feel that the central message is accurate – I give money to charity because I'd feel guilty if I didn't. To adopt the terminology used at Less Wrong, I don't buy (many) warm fuzzies with my donations. Instead, with each monthly donation I buy another month in which my conscience won't be gnawing at me in idle moments.
It was this ill-defined sense of guilt that got me started donating to charity in late 2008. I was approached by one of those charity representatives in the CBD, who I had usually ignored. But by then I had a full-time job lined up, and the excuses I'd been telling myself to avoid giving to charity – excuses which were already quite poor, as I was easily saving money on my postgrad scholarship and top-up – were now looking extremely weak. I agreed with the suggestion of giving $40 a month to Mission Australia for the next two years.
I don't know precisely how my thinking evolved over the next ten months. Certainly at some point early in 2009 I realised that a charity working in the Third World was probably doing much more cost-effective work than one working in Australia. In May, my now slightly-better-defined guilt became enough that I started donating to MSF, and I enjoyed the warm fuzzies I got from reading their monthly newsletter magazine.
By August, the rational part of my brain was trying to assert something of a utilitarian approach to giving, and I asked some MSF representatives if they had an idea of their organisation's overall cost per life saved (in some of the newsletters, a cost per life saved figure was given for particular vaccination programmes). I think I discovered GiveWell in September 2009. The idea that dollars I spent on myself could be saving people's lives became an enormous moral pull on me – I was statistically murdering innocent people – and I was soon donating to several of GiveWell's then-recommended charities. The guilt that led me to do so slowly crystallised into something resembling a Peter-Singer-style argument for why I should do so, and I sketch out this argument now. I've never actually read Singer in detail, so perhaps it's best to think of what follows as "what is in my head, in large part inspired by Peter Singer", rather than "Peter Singer's reasoning". I'm also influenced by David Boonin.
The classic Singer argument on this topic starts with the child drowning in a pond. You are passing by and have the ability to save the child, but it would wreck the expensive suit that you're wearing. Pretty much everyone agrees that you have a moral obligation to save the child. You have this moral obligation even if there are other adults by the pond, none of whom are willing to save the drowning child.
If it is possible to save a statistical life* with a donation to charity of the approximate size of the cost of an expensive suit, then you have a moral obligation to do so, just as you would have a moral obligation to save the drowning child.
*When you make your suit-sized donation, it doesn't directly save any particular person's life. But if you and enough others put in enough money, then the charity can expand its programmes – distributing bednets, say – to a new village. In the new village, more people will sleep under nets, and fewer people will die of malaria. It is possible to estimate how many fewer people will die, and what the total cost is, thus giving us a value for the cost needed to save a "statistical life".
Furthermore, almost everyone (everyone?) earning a high income by global standards today is lucky; for most of us this luck is being born into an affluent society. Through no fault of their own, millions of people are born into poverty with no opportunity of earning such high incomes. Taking a somewhat Rawlsian approach, if I didn't know where in the world's society I was going to be born, then I would want affluent people to see it as their moral duty to spare some of their cash and at least try to prevent people (possibly me) from dying of cheaply preventable diseases. In the "Where to give" section below, I describe why we can believe that saving such statistical lives is possible. For the moment, I will take it as an assumption.
The switch from saving a life directly to saving a statistical life is one which usually alters moral intuitions. I asked the following question to Facebook. It is similar to asking whether you have a moral obligation to donate to a high-impact charity, but re-cast so that it is you personally doing the work, rather than you outsourcing the work to an organisation in some country you've probably never been to.
You're driving through a city where no-one knows you. You notice a problem with the road. Due to a series of contrived circumstances (you know that the local government is dysfunctional, no-one in this city will believe you, etc.), you are the only person who might fix it. Luckily, you know what to do, and for $2000 you could quickly repair the road, and the repair would last for a year. You also know your traffic statistics. By repairing that bit of road, you reduce the probability of someone dying as they pass through it by 1 in 1 million. You also know that over a year, 1 million people will cross that bit of road. Do you have a moral obligation to spend the $2000 and reduce by one the expectation value of the number of road fatalities over the next year?
Everyone who answered said that their initial reaction was that they wouldn't save that statistical life. So when I say that you should be saving statistical lives by donating to charity, I expect that some part of your brain will agree with me, but that your moral intuitions will put up a kind of unexplained wall, saying "No, that can't be right."
Fortunately, the decision to donate to charity is one that you can make after careful thought. You are free to ignore your intuition if it is telling you that morally, it doesn't make much difference if you donate to a high-impact charity or not. I am unashamedly utilitarian on this point (and after a couple of years of trusting in reason, I now find that my moral intuitions agree with saving statistical lives). I think that you should also be utilitarian here – fewer people will die if we all donate to high-impact charities. That is why I give.
Where to give
Now that we are thinking with our utilitarian hats on, it should be clear that we want to give to the charity with the highest cost-effectiveness, defined as its total output divided by total revenue (for me, the output I'm interested in is saving statistical lives). Being more careful, we should define the cost-effectiveness of a charity as its marginal output divided by its marginal revenue – a charity might do wonderfully cost-effective work, but have already picked all the "low-hanging fruit" in its field, so that any extra dollars donated will go towards less cost-effective programmes.
The latter point is one regularly made by GiveWell. Before recommending a charity, GiveWell try to make sure that it has room for more funding for its high-impact programme(s). This isn't an obvious thing to look for, and it's an indication that finding the best charity is difficult.
More generally, there are several reasons why we should expect it to be hard to find good charities. These are spelt out in an excellent blog post by Holden Karnofsky (co-founder of GiveWell) – a good charity for a casual donor is working in an area which isn't already covered by the for-profit sector, governments, local charities and communities, and large foundations.
Still, one could point to all the poverty in the world, and all the deaths from malnutrition and malaria and so on, and say that there are obvious opportunities for doing large amounts of good with relatively small donations. And that is true. Ideally then, donors like me could go to charities' websites, study carefully their financials and their measured outputs (say, an increase in vaccination coverage compared to similar, control, areas where the charity isn't working), and be able to come up with an estimate of the cost per statistical life saved.
This is pretty much impossible. Financial reports are usually available, but detailed programme breakdowns can be hard to find, and measuring impacts relative to a control is extremely rare.
This is not a good situation, because the cost-effectiveness of different programmes can vary over several orders of magnitude (and some could potentially do net harm). The Disease Control Priorities Report (DCP2) has several graphs demonstrating this. See, for example, the graph on page 41 here; costs per DALY averted range from less than $10 to almost $100,000. As it happens, one of the particularly cheap health interventions mentioned in DCP2 – deworming – was found by GiveWell to be based on a calculation with several errors in it. The errors combined to increase the cost per DALY averted by two orders of magnitude. (And then, GiveWell's own calculation, taking other effects into account, reduced that enlarged cost-effectiveness estimate by a factor of about 5.)
So, most charities either don't measure or don't tell us their true impacts, and even when impacts are measured, the error bars can extend across orders of magnitude – what a mess! Fortunately, we are not completely helpless. As unsubtly hinted at above, GiveWell is an excellent resource for the casual donor. They start with a very high level of scepticism (perhaps too high), and try to find charities which transparently implement demonstrably cost-effective programmes and which have room for more funding.
The result is a very small number of recommendations – currently there are two top-rated charities, with a further six labelled "standout" (typically the latter are in fields other than developing-world health, and would be useful to donors who specifically want to contribute towards, say, microfinance or education). You can read detailed reviews of their recommended charities; here is their report on The Against Malaria Foundation, currently GiveWell's highest-rated charity.
The reports show an attention to detail that casual donors can't find elsewhere. One of many possible excerpts (edited to remove links to footnotes):
As of November 2011, AMF was just beginning to more systematically determine needs on a household-level basis. We have seen two examples of pre-distribution surveys, in which distribution partners determined how many sleeping spaces in each household were not covered by a useable insecticide-treated net and reported this data to AMF. We have seen details of how the survey was conducted for one of the two examples, and it appears to be of relatively high quality, while for the other we have only seen the completed beneficiary list and not the process that went into creating it. AMF told us that conducting pre-distribution surveys and creating beneficiary lists before the distribution will be a requirement for all distributions by March 2012.
Presumably there are some philanthropists who commission their own research into topics like these, but GiveWell is by far the best resource currently available for casual donors. You just can't find comprehensive detail on high-impact charities (both on the positive points and the areas where there's no currently known information) like this elsewhere. Their spreadsheet (XLS) used to estimate the cost-effectiveness of bednet distributions is wonderful – it goes well beyond the bednet-years per life saved figure reported in the academic literature, as it takes into account an empirical estimate of how long the nets are really used for, and (more importantly) pre-existing net coverage by demographic and hence the percentage of distributed nets that will go to those most at risk of dying from malaria. (For those interested, their base case marginal cost per life saved by bednet distributions is estimated at $1700.)
I don't want to overstate the confidence that you should have in GiveWell's recommendations – they have made revisions in the past, with (for instance) PSI and Stop TB Partnership being demoted for various reasons after GiveWell studied new information. If you want to estimate how many statistical lives you've saved, then it is perhaps best, after you've divided the total amount you've donated by the cost per life saved, to multiply by some value p (roughly the probability of an accurate calculation) less than 1.
But I don't want to understate the confidence that you should have in GiveWell's recommendations either. Subjectively, I would place that probability I mentioned comfortably above 0.5 – GiveWell's reports are consistently improving (the organisation is only about five years old). Speaking somewhat subjectively, I would add that especially for bednets, the case is very strong – for as long as I've been aware of the topic, I have heard people recommending bednet distributions as a very cost-effective way to prevent deaths from malaria.
More importantly, I think you can have more confidence in GiveWell's recommendations than in any others. The Poor Economics website lists recommended charities featured in the book by Banerjee and Duflo, but you only get a brief summary of each. Innovations for Poverty Action have more detailed, RCT-based recommendations, and they are perhaps the best alternative to GiveWell. If the latter didn't exist, I might be an IPA zealot instead of a GiveWell zealot. But I feel that GiveWell's reports show a level of sceptical analysis above what the IPA gives for its "Proven Impact" programmes, and so I have more confidence in the GiveWell recommendations.
Another group that needs mentioning here is Giving What We Can. Their recommended charities now overlap closely with GiveWell's (they recommend GiveWell's top two and one other). This is not coincidence – GWWC people and GiveWell people had many discussions last year, both privately and across several posts on the GiveWell blog. I don't know if it's coincidence that the Schistosomiasis Control Initiative was previously a GWWC recommended charity and is now also a GiveWell recommended charity; in any case, the fact that the two groups have come close to agreement, and that GWWC links to GiveWell's charity reviews, is further evidence that to maximise impact, we should follow GiveWell's recommendations.
A reasonable objection would be that we should give to GiveWell directly, so that they can optimally fund their research, and re-grant whatever money is left over to the top-rated charity they find. I don't have a strong counter to this. The best I can do is to say that while I trust GiveWell's recommendations more than any others, I like to have the option of sending my money to a charity which isn't their top-rated. More fundamentally, I think I just like the warm fuzzies I get from donating to the high-impact charities directly. (And if the AMF get their DGR status from the Australian Tax Office, then I'll get tax-deductibility from my donations to them. Usually the differences in charity effectiveness are far higher than my marginal tax rate, but in this case I think the taxes would win.)
Throughout the above discussion, I referred to GiveWell's recommendations in the plural. When maximising the expectation value of the impact of our donations, though, we should pick the one charity we believe has the highest expectation value of impact. The simplest thing to do is to donate to the #1-ranked charity (AMF) with no further thought. An alternative is to consider SCI (the #2) – which is better, one death due to malaria (and several hundred non-fatal cases of malaria) averted, or 250 children aged 5-14 dewormed for 10 years? That is roughly what you get for a $1700 donation, and it is up to the individual donor to consider how bad worm infections are before deciding on which intervention they believe has the highest impact.
There is no point being risk-averse and spreading out your donations across several charities. (Maybe if everyone donated the same way, this would be a problem. But GiveWell's money moved is tiny compared to the total amount people donate.) I did, for a fairly long time. Over September 2009, I increased my monthly MSF donation, and started giving to Village Reach, Stop TB, PSI, and The Global Fund. At the end of 2009 I decided to start giving to the AMF as well. As my moral intuitions became more and more utilitarian, this state of affairs became increasingly embarrassing to me (especially the Global Fund, who at best were about 40% the cost-effectiveness of the other then-GiveWell-recommendations.) It took till November 2010 to transfer my Global Fund donations to Stop TB, and till early 2011 to cancel my MSF donations.
Even today, I still split my donations. Despite GiveWell saying that VillageReach has essentially met its immediate funding needs, I didn't cancel. I figure that even if marginal dollars just generate interest for a while, VillageReach have such potential upside that it might be worth it (previously, GiveWell's report had estimated them saving lives at about $500 each). Of course, if I really believed VillageReach has the higher expectation value of impact per marginal dollar, then I should be giving exclusively to VillageReach. But I'm not perfect, and my moral intuition is not so finely honed into utilitarianism that it objects to having about 25% of my donations going to VillageReach and 75% to the AMF.
That might be an appropriate way to finish this section, but there are further issues that I think are worth discussing. There are negative impacts to foreign aid – the local currency appreciates, making life more difficult for their struggling exporters, and NGO's or non-profits can offer relatively high salaries, and hence might be diverting talented locals away from entrepreneurial work, thus potentially reducing economic growth. With a cost per life saved at around $2000, I think it would take a particularly optimistic view of future economic growth prospects in the absence of your marginal dollars, combined with a particularly pessimistic view of the negative impact of marginal dollars, to decide that you shouldn't save that statistical life.
Even if the gross amount of bad done by the marginal dollar is a dollar, and it actually costs $4000 to save a statistical life, then as long as you value a statistical life at more than $4000, the cost-benefit comparison says that you should donate. I value a statistical life much higher than that – perhaps something like 100 times the average annual income, as it is (roughly) in the West. That would equate to about $30,000 for someone living on a dollar a day.
Kremer et al. find a revealed preference value of a statistical life in Kenya at just $769. I am happy to be very paternalistic on this matter and carry on with my ball-park $30,000 figure. But someone who is very pessimistic about the effects of donations could use the revealed preference figure to justify not donating. Alternatively, someone who wishes to donate and not be paternalistic might be persuaded that an organisation giving out cash is preferable to one distributing bednets. Such a person could donate to GiveDirectly.
One last idea on "where to give", which can be carried out independently of everything I write above, is persuading others to switch their giving to high-impact charities, and/or lobbying for the government to redirect some of its official development assistance. (Probably doomed to failure is lobbying for an open-borders immigration policy, or an immigration tariff.)
The latter idea is one that I can't talk about with much authority. I know just enough about government-level foreign aid to be aware that I wouldn't have the slightest clue what to write in a letter to my local MP, or in a submission to an inquiry (if there even are inquiries with public submissions about Australia's ODA – I just don't know). But since AusAID has an annual budget of about $4 bn, there is probably a lot of scope for increasing the positive impact of Australia's ODA. Furthermore, foreign aid is not a topic which many people care a lot about, so it seems plausible to me that a well-informed citizen has a much higher probability of influencing government policy in this area as compared to, say, the health system. I don't think it would be easy, since the established position appears to be that ODA should be an extension of the Australian government's foreign policy (I recall news about increasing aid to Africa being derided as a blatant attempt by Kevin Rudd to curry favour with African governments in return for votes at the UN). But despite a low chance of success, it looks to me like lobbying the government has a high expectation value of positive impact after a moderately large effort.
When to give
I leave the bigger question of how much to give for last. Here I will take a quick detour on when to give, since it is a non-trivial problem with some interesting proposed answers. Not surprisingly, Holden has blogged on this topic, and his consideration is the most thoughtful I've seen.
Traditionally there are two approaches to take. One is that you give away your money regularly (be it weekly, monthly, annually) to whichever you think is the best charity. The alternative is to save and invest your money now, and give later. Being risk-neutral utilitarians, such investments should be in high-risk assets, so that the expectation value of the return is both the market return plus a risk premium. While your own assets may turn sour (it is risky, after all), a large number of utilitarians adopting this approach will spread out the risk and increase the total amount donated (when everyone eventually donates).
There are several defences you can make for the usual option of regular giving. The most obvious to me is the possibility that high-impact giving opportunities will disappear in the future. Many countries in sub-Saharan Africa have seen strong economic growth over the past decade, and if this continues, then vaccinations and bednets and the like will increasingly be able to be covered by the private sector and by local governments with larger tax bases.
In the linked blog post, Holden argues that following GiveWell's recommendations today means that GiveWell's funders will be happy to continue funding the operation. Over time, GiveWell's research will improve, and it is possible that future research will uncover better giving opportunities.
The more interesting defence, for my current purposes, is to argue that there is a "social return" on a donation, and that this return is greater than the return from investments in risky assets. I don't remotely know how to quantify such a social return; my rational brain thinks that while the social return exists (people able to work more, children attending school longer and earning larger incomes as an adult, and so on), it is probably smaller than the possible gains from risky investments. Holden's feeling is the opposite of mine, though, and I certainly feel like the social return is high.
So let's assume that a donation made now has a much higher return than what you can get from investing. (And let's also apply a discount rate, so that we can basically ignore the social return after a donation made only after a long-term investment.) This return will also be higher than the interest rate you'd get on a loan, so (as commenter Steve says in that blog post) the appropriate course of action is to borrow now to make a large donation today, and pay off the loan as you would a mortgage.
I don't know if it's even feasible to take out such a loan (if you default on your mortgage repayments, the bank can take your house; if you default on a loan that funded vaccines...?), but it seems to me that there are circumstances where it is appropriate. Suppose that GiveWell is moving $10 mn annually, and that their research uncovers an opportunity to save statistical lives very cheaply (say $200 each), and that this excellent charity can absorb another $20 mn. I think it would be optimal to have some donors borrow to make large donations now, so that this excellent work is fully funded immediately.
I very much doubt that anyone would do this (and I'm not even convinced that the social return is high enough to justify it). In the meantime, I will continue giving monthly, because I want to give now and because monthly donations are relatively small and don't require fighting with the bank to increase the amount of money I can transfer.
How much to give
'Jesus sat down opposite the place where the offerings were put and watched the crowd putting their money into the temple treasury. Many rich people threw in large amounts. But a poor widow came and put in two very small copper coins, worth only a few cents. Calling his disciples to him, Jesus said, "Truly I tell you, this poor widow has put more into the treasury than all the others. They all gave out of their wealth; but she, out of her poverty, put in everything—all she had to live on."' – Mark 12:41-44 (NIV)
It may be obvious to you that there's no clear answer on how much you should give. But I agonised for months over it. On the one hand, my mind repeatedly refers back to the lesson of the widow's mite. On the other hand, it is clear that large donations do more good than smaller ones, so it is morally wrong to deliberately earn a small income so that you maximise your "widow's-mite-style" giving.
I therefore propose the following metric for how "morally good" you are: it is the expectation value of the impact of the donations you make, divided by the expectation value of the impact of the maximum potential donations you could make. The denominator would be achieved when you maximise your salary (I might have to be a hyper-ambitious banker), minimise your living expenses (I'd have to move in to share accommodation, cancel TV and Internet subscriptions, stop eating at restaurants, etc.), and donate your entire discretionary income to the charity with the highest expectation value of impact.
(The above applies to regular wage-earners. To be slightly more general, the denominator involves maximising the expectation value of your income. For people with some entrepreneurial skill, this may involve founding a start-up. Hall and Woodward find that the average amount of exit cash received by an entrepreneur who gets venture capital funding (this is difficult) is $5.8 mn. Most startups fail, but some become the Googles or Facebooks, and we should be risk-neutral in our utilitarianism.)
That is one extreme. Such a life would be an extremely onerous one, so it is sensible to step back from pure utilitarianism and find a more attainable standard. The logical moral requirement would be that you should donate at least your "fair share". How would such a fair share be defined? It might be possible to add up the total amount of funding required for all very-high- and high-impact interventions, and spread this estimated amount across almost all the income-earners in affluent countries. A similar idea, resting on contestable assumptions, underlies Peter Singer's suggested scale, which progresses from requiring 1% of income to be donated to charity by relatively low income-earners, to one third for the top 0.01%.
I would put a minimum requirement for a fair share at 0.7% of income, both because it feels like a reasonable lower bound, and it is already something of a standard.
Before continuing, I will make a little diversion on taxes. If you believe, as I do, that almost every income earner has a moral obligation to donate to a high-impact charity, then it is reasonable to conclude that such giving should be forced through the tax system so that there are no shirkers. You could argue that it would fulfil your moral duty to vote for a political party which promises to expand the foreign aid budget and have it focus on transparent, high-impact interventions. I think that such behaviour is only sufficient when there is a decent chance of such a party actually winning government, and that is not the case. By contrast, there is a large constituency in favour of the welfare state, and so any moral duty to help less well-off fellow citizens or residents might be fulfilled by voting accordingly.
So I see your moral duty as giving at least 0.7% of your income to a high-impact charity. I then think it is up to the individual's conscience to decide how utilitarian they should become, in approaching the maximum-discretionary-income donations.
My own conscience says that I need to give much more than just 0.7%. Recall the drowning child when there are several adults standing around idly – you still have a moral obligation to rescue the child. That is the situation we have today, with high-impact interventions still under-funded. Every $2000 I spend on myself (or fail to earn and donate) is a statistical murder.
I don't usually think in terms of myself as a statistical murderer these days, but I did in 2009. Nevertheless, I found it difficult to make the leap to donating substantial sums. Before I discovered GiveWell, I resolved to increase my monthly MSF donations from $80 to $250; on the phone I was too nervous and said $200 instead. My conscience told me off, but I hesitated because when I knew no-one who was giving away all their discretionary income. The evidence of the behaviour of the rest of society made me doubt my utilitarian reasoning.
I didn't continue to doubt it for long, but I gave in to my conscience gradually. You may recall the array of charities I started donating to in September 2009. It started with the $200 MSF donations. Then, after discovering GiveWell, I set up a monthly donation to VillageReach of 160 USD. A couple of days later, my conscience wasn't satisfied. I set up a 200 USD monthly donation to Stop TB. Then $200 to PSI. A few days later I still wasn't happy, and set up a $200 monthly donation to the Global Fund, and upped my VillageReach donations to $300.
I don't know precisely when I learned about Zell Kravinsky. It would make sense if it were around this time, but all I know is that I was using him as a rolemodel by early 2010. Kravinsky made a fortune in real estate, building up his wealth to $45 mn. He then gave almost all of it away (and a kidney as well). It was a relief to learn about Kravinsky – an affirmation that I was correct in believing that I should be giving much, much more than I had been before September 2009.
Those donations kept me satisfied temporarily. I was (and still am) earning Australian dollars, and if the AUD weakened to its 2009 low of 65 US cents, then the monthly USD donations would leave me with no spare money. But by December, the currency was holding firm (and the AUD has since appreciated by about 15% relative to the USD), and so I started making irregularly-sized monthly donations to AMF in AUD. The first was an arbitrarily-sized $1000; subsequently I decided to almost cease putting money into my savings, and gave whatever was needed at the end of the month to have spent my month's salary. Over the calendar year of 2010, these donations averaged $800 per month.
My monthly after-tax salary was then about $3900, and I was donating about $1900. This level of donations satisfied my conscience, although the sub-optimal array of charities gnawed at me. I didn't (and still don't) feel a moral requirement to undergo retraining and try to make it big in the finance industry; my conscience is also happy with me living on my own, paying substantially more rent than I would in a sharehouse. I'm a long way from perfectly utilitarian.
For a time in 2011, there appeared a sizeable chance that I would need to spend a decent amount of time in hospital. While a part of my brain idly wondered what the correct utilitarian approach was, in practice I quickly cancelled those embarrassing MSF donations and paused the AMF donations, with no qualms (I further "rationalised" my donations mid-year, when my credit card expired). As it happened, I didn't need anything more than a day procedure, and my savings built up nicely. In September I resumed the AMF donations, though I was still anchored to the idea of building up my savings, so the donations were a few hundred dollars short of what they could have been. In January I resumed full donations, with 600 USD going to VillageReach and $2000 to AMF. That comes from an after-tax salary of $4400. I suspect that the $2000 is unusually high, and that I'll settle in at donating about half of my after-tax salary.
I don't expect any of you to immediately start donating half your salary (especially if you have kids, which I don't). If you're a non-donor and have made it this far, though, then I would recommend trying to give the traditional 10% tithe. I chose 10% there because that is the target set by Giving What We Can, and, especially considering its history in the church, it has as good a chance as any target of becoming a social norm. It's also quite achievable and would (depending on your salary) result in saving many statistical lives over the next few years.
In an ideal world, it wouldn't be possible to prevent people dying at such low cost, as the interventions would have already been paid for. But the world is far from ideal, and I believe that those of us who recognise this have a moral obligation to do a substantial amount to rectify it.